mb706

Try to solve some issues under the [POFU Tag](https://github.com/mlr-org/mlr3pipelines/issues?q=is%3Aissue+is%3Aopen+label%3A%22Tag%3A+POFU%22)

Use DataBackend info to avoid unnecessary data comparison in #271

1

#271 solution currently needs to compare data to make sure clashing columns are not different. Once https://github.com/mlr-org/mlr3/issues/476 has an answer we can speed up this test.

PipeOpFeatureUnion could drop all targets except first task's one

1

Instead of checking that all targets are the same and throwing an error if they are not.

PipeOpFeatureUnion with differing row IDs?

2

PipeOpFeatureUnion could under some circumstances want to unite tasks that have differing row IDs, e.g. after PipeOpSubsample on two different paths sampled (and `$filter()`ed) different sets of rows. ```r graph...

see if tests pass with proposed changes in paradox

trying a universal repr system for mlr3 and friends Related: * https://github.com/mlr-org/mlr3misc/pull/34 * https://github.com/mlr-org/mlr3/pull/384

move POBackupLearner out of attic

1

Low priority, if this is ever shown to have benefits over the mlr3-internal functionality. This reverts commit 58fb1d022b46470994520fe50558e99f1b8ac083.

From here: https://github.com/mlr-org/miesmuschel/blob/smashy_ex/R/PipeOpPredictionUnion.R I think we already have something similar somewhere else.

Mostly the problem that occurred here: https://github.com/mlr-org/miesmuschel/blob/smashy_ex/R/PipeOpStratify.R The necessity for a fallback should be visible automatically, somehow.

Change documentation to roxygen R6 style

1

We will probably have to do this sooner or later, although the style may not fit well to document individual PipeOps.